Mario Frasca writes:
>
> Hi Al,
> I think this will be interesting for most people on the list...
> well, I've written this version of init.c and modified date.c.
>
> *** init.c does a decent job, according to me, allowing the elks
> administrator to write a /etc/inittab file, but it sometimes (quite
> regularly) corrupts utmp. I fear that I'm using too much memory, so I'm
> rewriting it again (but I have much less time now, compared to the
> previous weeks), this time I only keep in memory the ID and pid of
> running children which need to be respawned. I will not care about
> anything else. Do we have anything to measure the memory we still have
> at disposition? so I can check any possible memory leak and compare
> memory usage. the problem with the utmp is not a direct issue to me, so
> I prefer focusing on memory usage first.
We don't have anything at the moment, but you can always stick a printk in
kernel space so it prints the ammount of memory used by init periodically.
By default programs pre-allocate at least 32K and all the memory they
require must come out of this. You can reduce this value at compile time.
>
> *** date.c allows you to *insert* the current date (only ISO format
> [yy]yy-M-d[Th[:m[:s]]] is understood). there are three ways to do this:
>
> date <date> (date is given as parameter)
> date ? (stop and ask for the current date)
> date ?? (checks current date, if before 1992, it asks for the current
> date)
Okay, these make sense.
> a date without century is interpreted as 1970-2069.
This is patented. It is not legal to use this algorithm to deal with y2k
issues. See this link for more info:
http://slashdot.org/articles/99/11/01/2047228.shtml ;-)
>
> I included 'clock -s ; date ??' in my /etc/rc.d/rc.sysinit, specified
> in my /etc/inittab as
> si::sysinit:/etc/rc.d/rc.sysinit
>
> oh, yes:
> under elks 0.0.79 this gave some problems, since only *one* virtual
> console was spawned (I had to log in as root and give 'init <current
> runlevel>' to start them). in 0.0.81 this has been solved.
> under 0.0.81, clock does not work any more on my PPC640: it reads
> rubbish from the CMOS clock and writes such rubbish that after that the
> computer thinks the CMOS clock is broken. so I took 'clock -s' away
> from my system initialization.
>
> how do I contribute the two programs? is 'just mail to the
> list' an option?
>
Mailing to the list is fine.
Al